rosinality/swapping-autoencoder-pytorch

Minor Issue

bryandlee opened this issue · 1 comments

Hi, thanks for the great work.

There is a typo in the pad size computation formula for the upsample blur kernel:

p = (len(blur_kernel) - factor) + (kernel_size - 1)

I guess the sign should be changed

p = (len(blur_kernel) - factor) - (kernel_size - 1)

Thank you, fixed it at 0096ec7.